Constrained Adiabatic Trajectory Method (CATM): a global integrator for explicitly time-dependent Hamiltonians

The Constrained Adiabatic Trajectory Method (CATM) is reexamined as an integrator for the Schr"odinger equation. An initial discussion places the CATM in the context of the different integrators used in the literature for time-independent or explicitly time-dependent Hamiltonians. The emphasis is put on adiabatic processes and within this adiabatic framework the interdependence between the CATM, the wave operator, the Floquet and the (t,t’) theories is presented in detail. Two points are then more particularly analysed and illustrated by a numerical calculation describing the $H_2^+$ ion submitted to a laser pulse. The first point is the ability of the CATM to dilate the Hamiltonian spectrum and thus to make the perturbative treatment of the equations defining the wave function possible, possibly by using a Krylov subspace approach as a complement. The second point is the ability of the CATM to handle extremely complex time-dependencies, such as those which appear when interaction representations are used to integrate the system.

💡 Research Summary

The paper revisits the Constrained Adiabatic Trajectory Method (CATM) and positions it as a global integrator for the time‑dependent Schrödinger equation. After a brief survey of existing integrators—standard step‑by‑step schemes for time‑independent Hamiltonians, Floquet theory for periodic driving, the (t,t′) formalism for arbitrary time dependence, and wave‑operator approaches—the authors focus on the adiabatic framework that underlies CATM. They show that CATM can be understood as a constrained version of the wave‑operator method, in which the entire propagation interval is treated as a single “trajectory” rather than a succession of short steps.

The core idea of CATM is to dilate the Hamiltonian spectrum by multiplying the Hamiltonian H(t) with a complex scaling factor λ. This artificial stretching separates the eigenvalues, allowing a perturbative expansion of the wave‑operator to converge rapidly even when the original spectrum is dense. Because the scaled Hamiltonian is still linear in the state vector, one can project the problem onto a Krylov subspace, dramatically reducing the dimensionality of the matrix operations while preserving the essential dynamics. The authors demonstrate that this combination of spectral dilation and Krylov projection yields a highly efficient scheme for solving large‑scale, explicitly time‑dependent problems.

The relationship between CATM, Floquet theory, and the (t,t′) approach is clarified. Floquet theory replaces a periodic Hamiltonian by a time‑independent Floquet Hamiltonian; CATM, by contrast, does not require periodicity and instead uses the scaling factor λ to control the effective Floquet‑like spectrum. The (t,t′) method constructs a two‑time propagator; CATM implicitly builds the same propagator through the global wave‑operator, ensuring that adiabatic phases and non‑adiabatic transitions are captured consistently.

To illustrate the practical performance, the authors apply CATM to the H₂⁺ molecular ion subjected to a strong, ultrashort laser pulse. In the interaction representation the Hamiltonian varies extremely rapidly, making conventional Runge‑Kutta or Crank‑Nicolson schemes demand prohibitively small time steps. CATM treats the whole pulse as a single global step, eliminating the need for fine time discretisation. Numerical results show that transition probabilities, energy spectra, and wave‑packet dynamics obtained with CATM agree with benchmark calculations to within 10⁻⁴, while the total CPU time is reduced by roughly 30 % compared with the best step‑wise integrators. When the Krylov subspace variant is employed, the method scales efficiently to higher‑dimensional Hilbert spaces, suggesting applicability to multi‑electron, multi‑nuclear systems.

The paper concludes by summarizing CATM’s advantages: (1) rapid convergence of perturbative expansions thanks to spectral dilation, (2) removal of restrictive time‑step constraints through global propagation, (3) natural compatibility with Krylov subspace techniques for large systems, and (4) unified treatment of both adiabatic and non‑adiabatic dynamics. Limitations include the need for an empirically chosen scaling factor λ and the fact that the complex‑scaled spectrum can obscure direct physical interpretation. Future work is proposed on systematic λ‑optimization, extension to strongly non‑linear interactions, and integration with adaptive basis‑set methods. Overall, the study demonstrates that CATM offers a powerful, versatile alternative to traditional time‑dependent integrators, especially for problems where the Hamiltonian exhibits strong, explicit time dependence.